A case study of early impedance spectroscopy for corrosion forecasting in reinforced mortar

This case study examines whether early electrochemical impedance spectroscopy (EIS) adds forecasting information beyond an available linear-polarisation-resistance (LPR) measurement in reinforced fly-ash and fly-ash–metakaolin mortar. A 90-specimen wet–dry exposure study provided 332 reconstructed LPR-derived current-density labels. Separate reconstructed and archived label conventions defined cycle-7 landmark cohorts of 56 and 66 specimens, with cycle-49 log-current as the endpoint. Elastic-net and partial-least-squares models were compared using matched nested validation, condition-cell bootstrap intervals, calibration analysis and ten fully retrained condition-group partitions. Under reconstructed labels and replicate holdout, adding full EIS to conditions and early LPR reduced mean absolute error (MAE) from 0.482 to 0.417 for elastic net and from 0.470 to 0.421 for partial least squares. The corresponding conditional MAE differences were −0.064 (95% interval −0.120 to −0.009) and −0.049 (−0.104 to 0.003). However, archived-label point estimates did not improve, and mean differences across retrained condition-group partitions were only −0.002 and −0.010 under reconstructed labels. Exposure-solution holdout increased full-spectrum MAE to approximately 0.85–0.86 and produced poor calibration. The results are consistent with impedance reflecting both matrix/exposure state and interfacial processes, rather than a unique future-corrosion signal. They support label-consistent, baseline-referenced assessment of laboratory monitoring data, but do not establish a transferable EIS advantage or a validated corrosion-warning system.